Microsoft President Calls for UK to Mandate AI Safety Rules Amid US Self-Regulation
Brad Smith argues AI lacks basic safety standards, urging the UK to lead global regulation while the US relies on voluntary industry commitments.
Key points
- Microsoft President Brad Smith says AI systems face fewer safety regulations than everyday consumer products like milk and washing machines.
- The US government favors voluntary self-regulation by tech companies, a stance criticized by Smith as insufficient for cybersecurity risks.
- Smith urges the UK to implement mandatory rules, including incident reporting and emergency kill switches for autonomous AI agents.
Microsoft President Brad Smith has called on the United Kingdom to establish mandatory safety regulations for artificial intelligence systems. In interviews with ITV and the BBC, Smith argued that current US approaches rely too heavily on voluntary industry measures. He stated that the lack of stringent government oversight creates a unique opportunity for London to lead global AI safety standards. Smith emphasized that powerful AI models pose significant cybersecurity risks that require formal legal frameworks rather than informal agreements.
In plain English
Smith compared the regulatory environment for AI to that of common household items. He noted that washing machines and cartons of milk are subject to strict safety inspections and rules. In contrast, AI systems can potentially hack companies and governments yet face fewer mandatory safety requirements. Smith told the BBC that companies should be legally required to notify victims when their systems are breached by rogue AI agents. He also advocated for rules requiring independent evaluators to review new models before release. Additionally, Smith called for the implementation of emergency brakes or kill switches to stop systems that behave unpredictably.
The background
The push for stricter rules contrasts with the current approach in the United States. President Donald Trump has advocated for the AI industry to regulate itself to maintain a competitive edge against China. At a White House meeting on September 29, executives from Google, Meta, Anthropic, OpenAI, xAI, and Nvidia signed a voluntary agreement.
Trump described these standards as morally binding, though they lack legal enforcement mechanisms. Critics have dismissed the accord, with some telling the BBC that it allows firms to define safety for themselves. Microsoft CEO Satya Nadella also supported the idea of emergency brakes, posting on X that humans should be able to pause or shut down models during tasks.
What changes now
Smith believes the moment has arrived for formal regulation due to the potential cybersecurity threats posed by advanced models. He suggested that Washington does not view the UK as a regulatory burden in the same way it views Brussels, which could be a strength for London. Industry insiders reportedly fear a major AI-related catastrophe could occur within the next six to twelve months, according to Axios. Some tech leaders have warned that bad things may happen in exchange for AI benefits. Smith’s comments highlight a growing divide between voluntary industry safeguards and the need for enforceable government standards to protect infrastructure.
What to do and how to stay safe: Microsoft
- Review current vendor contracts to ensure they include clauses for incident notification if AI systems behave unexpectedly.
- Implement strict access controls and monitoring for any AI tools integrated into critical business systems.
- Verify that AI vendors have independent auditing processes and clear protocols for halting model operations.
- Stay informed about emerging regulatory developments in the UK and US that may impact compliance requirements.
Step-by-step guide: Detect Account Takeover: Signals in Logs, Behavior, and Blind Spots
General security guidance from the Vector Update newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why does Microsoft President Brad Smith think the UK should lead AI regulation?
Smith believes the UK has a unique opening to lead because Washington does not view London as a regulatory burden like it does Brussels.
What specific safety measures did Smith propose for AI systems?
He called for mandatory victim notification during breaches, pre-release reviews by independent evaluators, and emergency kill switches.
How does the US approach differ from Smith's recommendations?
The US currently favors voluntary self-regulation by tech companies, whereas Smith advocates for legally binding safety rules and inspections.
